Docstoc

Anomalies as a Catalyst for Middle School Students’ Knowledge

Document Sample
Anomalies as a Catalyst for Middle School Students’ Knowledge Powered By Docstoc
					Journal of Educational Psychology 2003, Vol. 95, No. 2, 357–374

Copyright 2003 by the American Psychological Association, Inc. 0022-0663/03/$12.00 DOI: 10.1037/0022-0663.95.2.357

Anomalies as a Catalyst for Middle School Students’ Knowledge Construction and Scientific Reasoning During Science Inquiry
Marissa Echevarria
State University of New York at Albany
Knowledge construction and scientific reasoning of 7th-grade students were examined during a 3-week inquiry unit in genetics, in which anomalies were used as a catalyst for student learning. Students used genetics simulation software to develop hypotheses and run tests of fruit fly crosses in order to develop mental models of simple dominance trait transmission. Instruction was intended to support discovery of anomalous patterns and development of explanations. Qualitative and quantitative analyses indicated that student explanations showed a significant shift toward greater explanatory power of the anomalous inheritance patterns. However, this shift did not occur evenly. Students were more likely to propose hypotheses and explanations for the more frequently occurring anomaly and more likely to run the test that produced that outcome relative to the less frequently occurring anomaly.

Anomalous events encountered by the learner are widely considered to be a catalyst for conceptual change learning (Posner, Strike, Hewson, & Gertzog, 1982). As such, exposure to anomalous events has often been used in the teaching and learning of science to stimulate construction of science ideas. Use of this approach typically involves presenting an anomalous situation to learners and examining how their ideas might change. For example, learners might be presented with texts in which anomalous claims are examined in relation to existing theories (Limon & Carretero, 1997) or texts in which anomalous information is embedded in order to elicit questions (Graesser & McMahen, 1993). Alternatively, learners might be presented with actual demonstrations of anomalous events when dealing with physical phenomena such as buoyancy of objects in water (Burbules & Linn, 1988). What this approach indicates is that learners vary in their recognition of, and tolerance for, anomalies as well as the extent to which their ideas may change. However, what this approach does not illuminate is how the learner may further choose to investigate an anomalous event and what impact that investigation may have on the knowledge that is subsequently constructed. For example, in the area of science inquiry where students can develop hypotheses and design tests to investigate a phenomenon, will learners choose to pursue tests that produce anomalous outcomes? Or will they ignore those tests in favor of those that produce outcomes consistent with their theories? How will these testing approaches influence the knowledge that is subsequently constructed?

Marissa Echevarria, Department of Educational and Counseling Psychology, State University of New York at Albany. I thank Linda Gellman and students for their gracious participation in this project. This article is based on a doctoral dissertation submitted to the State University of New York at Albany. Portions of this article were presented at the annual meeting of the American Educational Research Association, Seattle, Washington, April 2001. Correspondence concerning this article should be addressed to Marissa Echevarria, who is now at the Graduate School of Education, 1207 Sproul Hall, University of California, Riverside, California 92521. E-mail: marissa.echevarria@ucr.edu 357

Research on scientific reasoning also does not provide an answer to these questions. This body of research focuses on various aspects of the processes involved in scientific reasoning, such as developing hypotheses, planning tests, conducting tests, and drawing conclusions. A typical approach in this line of research is to examine differences in these processes between experts and novices in a domain or between adults and children (Klahr, Fay, & Dunbar, 1993; Kuhn, Garcia-Mila, Zohar, & Andersen, 1995; Schauble, 1996). A common framework is to provide participants with a system to explore in which they must determine the causality or noncausality of several variables in relation to an outcome. For example, Schauble (1990) investigated the extent to which children were able to determine the relation between various design features of a race car and the speed of that race car. Using such a framework, investigators examined strategies of experimentation with respect to systematicity of testing approaches and validity of inferences drawn, in particular with regard to whether participants will control for variables during experimentation by only varying one thing at a time (Friedler, Nachmias, & Linn, 1990; Vollmeyer, Burns, & Holyoak, 1996). Although these studies are informative about approaches to scientific reasoning, they are not informative about subsequent changes in knowledge beyond that of establishing which variables a learner believes to be causal or noncausal due to patterns of covariation or noncovariation, respectively. For example, a learner may find that a toy boat moves slower in shallow water than in deep water but not know why such an outcome occurs (Schauble, 1996). Thus, there has been little study of the mechanism by which a learner believes a variable to be causally related to an observed outcome. Therefore, an apparent dichotomy exists in the research literature with studies on student ideas and explanations in response to presentation of anomalous information or on proficiency of scientific reasoning processes. This dichotomy holds for studies conducted in laboratory settings (Kuhn et al., 1995) as well as those conducted in the classroom (Roth & Roychoudhury, 1993; Shepardson & Moje, 1999). Knowledge construction and scientific reasoning are infrequently studied together to determine how processes of scientific reasoning in response to anomalies contribute

358

ECHEVARRIA

to the ideas and explanations constructed by the learner to account for anomalous phenomena. Yet this is the process engaged in by scientists when they encounter anomalous phenomena and then must choose how to conduct their investigations to further explore such phenomena. Thus, a more integrated approach to examining learner responses to anomalies might provide a more holistic and naturalistic look at how learners respond to anomalies, revealing the influence on scientific reasoning and construction of knowledge. The purpose of the present study was to examine knowledge construction and scientific reasoning of seventh-grade students during a 3-week inquiry unit on Mendelian genetics. During this unit, student responses were examined in relation to anomalous inheritance patterns that arose when offspring were produced that did not resemble the parents or that resembled only one of the parents. For example, for the trait of eye color (either red or white) in fruit flies, two red-eyed fruit flies might produce both red- and white-eyed offspring. This pattern was considered anomalous in that it was not apparent why white-eyed offspring would be produced from two red-eyed parents. In another case, a red- and white-eyed fruit fly might produce only red-eyed offspring. This pattern was considered anomalous in that no white-eyed offspring were produced from a white-eyed parent. During the unit, students used genetics simulation software to investigate the inheritance patterns of the fruit flies. Using this software, students could construct their own hypotheses, run tests of those hypotheses by selecting two fruit flies and mating them, observe the offspring produced, and draw conclusions. The goal of the unit was for students to develop a mental model of trait transmission based on their data. Change in student content knowledge was gauged through a preunit and postunit assessment that specifically focused on explanations of the two anomalous inheritance patterns. Approaches to scientific reasoning were examined on the basis of the number and types of hypotheses students generated that addressed anomalous outcomes and the number and types of tests students ran that generated anomalous outcomes.

Presentation of Anomalies and Construction of Knowledge
When presented with anomalies, learners may construct knowledge in various ways. For example, knowledge construction has been described as involving weak restructuring, in which small changes are made to an existing theory, or radical restructuring, in which a paradigm shift occurs in how the learner views the phenomena (Vosniadou & Brewer, 1987). Knowledge construction may also occur unevenly, evident in some situations, but not in others. Tao and Gunstone (1999) found that student conceptions during an inquiry unit in physics vacillated between canonical and alternative explanations depending on the context. Given the same problem, some contexts invoked canonical explanations, whereas others invoked alternative explanations. Demastes, Good, and Peebles (1996) described four different types of knowledge change when learning concepts of evolution. These processes are cascade, wholesale, incremental, and dual construction. Cascade involves a change in a key concept that allows other concepts to change. Wholesale involves completely discarding one theory for another. Incremental involves adopting one particular concept and gradually elaborating on it until it is fully integrated into the learner’s

prior knowledge. Dual construction involves maintaining two theories about the same phenomena without the learner being aware of their incompatibility. Similar patterns to wholesale change, dual construction, and incremental change have been reported elsewhere (Burbules & Linn, 1988; Vosniadou & Brewer, 1992). There are various reasons why change in knowledge may occur in a nonlinear fashion. Learners who maintain two theories to explain the same phenomena may not have an epistemological commitment to integrated, parsimonious explanations (Posner et al., 1982). Learners may also incrementally change conceptions as more instances of anomalous data can be explained, rather than relinquish a theory that explains some of the data in favor of having no theory to explain the data (Koslowski, 1996). Alternatively, when anomalous data cannot be explained, learners may respond in various “theory-preserving” ways to maintain their current theory. These responses include (a) ignoring the data without bothering to explain it; (b) rejecting the data on the basis of methodological or other grounds; (c) excluding the data as not being relevant to the theory in question; (d) holding the data in abeyance, confident that at some point the current theory would be able to explain them; (e) reinterpreting the data to be consonant with the current theory; and (f) making a peripheral theory change while leaving the core theory intact (Chinn & Brewer, 1993). On the basis of the conceptual change model proposed by Posner et al. (1982), a condition under which anomalous data are more likely to prompt conceptual change is when they are both intelligible and plausible. Data that are neither intelligible nor plausible will most often be ignored. Data that are intelligible but implausible are more difficult to assimilate and accommodate relative to plausible data. Thus, the learner may try to explain such data before accepting them in an attempt to provide a plausible mechanism. If no plausible mechanism can be posited, the learner may discount the data (Kuhn et al., 1995). When no mechanism information is present, implausible data may be viewed as more plausible when many instances of such data occur. Responses such as ignoring or rejecting anomalous data become increasingly untenable in the face of mounting evidence to the contrary. Koslowski (1996) reported that four instances of anomalous data, but not three, were found to increase causal ratings of a target variable relative to only one instance of anomalous data. Burbules and Linn (1988) contended that repeated presentations of contradictory information caused students to gradually change their conceptions regarding displacement of objects in water from a weight-based rule to a volume-based rule. However, students varied considerably in their tolerance for anomalous instances of data, with some students immediately changing their theory while others persisted. These researchers concluded that students seem to learn from a summative effect of repeated contradictions but that these contradictions may vary in their influence from student to student. The anomalies in the present study were considered to be intelligible to the learners but not necessarily plausible. Therefore, explanations were focused on during the unit in an effort to help students construct mechanisms that might help to render these implausible outcomes plausible. The extent to which students encountered the anomalous phenomena was dependent on student testing. Thus, this was a variable of interest with respect to determining whether students would pursue or eschew further tests that might produce anomalous outcomes.

KNOWLEDGE CONSTRUCTION AND SCIENTIFIC REASONING

359

Experts and Novices in Scientific Reasoning
Researchers who have examined student scientific reasoning have infrequently examined it in relation to anomalies. Rather, research on scientific reasoning has largely focused on processes such as hypothesis generation, systematicity of testing, and drawing of inferences based on an expert–novice paradigm. Domainspecific knowledge can play a large role in how proficiently these processes are executed. Specifically, domain-specific knowledge can aid in generating initial as well as alternative hypotheses (Klahr & Dunbar, 1988; Slack & Stewart, 1990). However, if the learner does not possess prior knowledge from which to draw, data may be gathered through experimentation in order to formulate a hypothesis (Slack & Stewart, 1990). When approaching experimentation, novices in a domain are more likely to haphazardly conduct an investigation relative to experts (Simmons & Lunetta, 1993). However, compensating factors when domain-specific knowledge is lacking are powerful domain-general strategies that can effectively equalize performance between novices and experts (Smith & Good, 1984). Producing controlled tests can also be related to the level of domain-specific knowledge. Friedler et al. (1990) found that students planned more controlled tests in a task with familiar variables compared with a task with unfamiliar variables, despite the fact that more variables needed to be controlled in the familiar task. They suggest that the informationprocessing load of dealing with familiar variables was less than that for dealing with unfamiliar variables. Hence, more information-processing capacity was available for constructing controlled tests with the familiar variables compared with the unfamiliar variables. The drawing of inferences involves interpretation of evidence to form a conclusion regarding the relationship between variables and an outcome. Differences in validity of inferences are found between novices and experts in a domain as well as between children and adults. Novices make fewer valid inferences than experts (Simmons & Lunetta, 1993) and may make unwarranted inferences given the data (Slack & Stewart, 1990). Children are also likely to make fewer valid inferences than adults (Kuhn et al., 1995; Schauble, 1996). With respect to types of inferences, both children and adults are more likely to draw inferences of inclusion or causality relative to inferences of exclusion or noncausality. Interaction inferences are generally difficult for learners to draw. This pattern involves coordinating multiple instances of data to ascertain an overall pattern. Learners are much more likely to draw simple inferences of causality or noncausality rather than to detect an interaction pattern (Kuhn et al., 1995; Schauble, 1996). In the present study, students were assumed to be novices in genetics and therefore were expected to conduct their tests somewhat unsystematically. Attention was paid to this fact during the unit through suggestions for how to approach testing. The genetic data that students were working with involved an interaction mechanism, rather than a simple causal or noncausal mechanism. That is, the genetic makeup of both parents interacted to produce offspring, rather than one parent creating offspring and the other parent playing a noncausal role. On the basis of the extant literature, it was expected that this type of inference would be difficult to draw and that it was more likely that students would attempt to draw inferences of inclusion or exclusion before interaction inferences.

Problem Solving in Genetics
In prior studies of genetics, two types of approaches to problem solving have been studied: (a) cause-to-effect reasoning and (b) effect-to-cause reasoning. For cause-to-effect reasoning, students are provided with the genetic makeup of the parents and asked to determine the number, type, and genetic makeup of the offspring that would be produced. For effect-to-cause reasoning, students are provided with the number and type of the offspring and expected to infer the genetic makeup of the parents. Cause-to-effect approaches to problem solving in genetics have often been associated with rote algorithmic practices on the part of students (Browning & Lehman, 1988; Smith & Good, 1984; Stewart, 1983). Using this form of problem solving, students may rely on memorization and recall of prescribed steps without an underlying conceptual understanding of what those steps represent. Effect-to-cause problem solving is less susceptible to rote practices because it is generally not taught in an algorithmic fashion. This was the approach used in the present study. A typical format that has been used with effect-to-cause problem solving is to teach students about a particular model of genetics and then to have them use that model to explain the effects they see (i.e., model using). An alternative approach is to teach students about a particular model of genetics and then to have them use that model as a basis from which to develop a more complex model (i.e., model revising). This practice is most closely associated with the practices engaged in by scientists. In studies on model revising, students have been provided with a baseline simple-dominance Mendelian model of genetics and then were given genetic patterns that did not fit the baseline model (Finkel, 1996; Hafner & Stewart, 1995). Students then had to revise the baseline model to explain the new genetic patterns. Students used different types of knowledge to conduct their investigations, such as genetics knowledge, knowledge of the process of model revision, and metacognitive knowledge of problem-solving strategies (Finkel, 1996). Students also used various heuristics such as conducting a systematic search of the experiment space, using existing models as templates to interpret new data, revising models to accommodate the new data, and using the new models to reanalyze the experiment space (Hafner & Stewart, 1995). Novices in genetics were likely to make unwarranted inferences from the data and to conduct their investigations in an unsystematic manner (Simmons & Lunetta, 1993; Slack & Stewart, 1990). Further, lack of robust domain-specific knowledge contributed to erroneous conclusions regarding the concept of dominance for a simple dominance problem (Slack & Stewart, 1990). Specifically, students assumed that the most frequently appearing form of the trait was dominant. They also concluded that the dominant form of the trait could change from one cross to the next and that if all of one type of offspring was produced, then that form was dominant. When encountering anomalies, students often ignored them or considered them to be mistakes or mutations. The format of the present study was also an effect-to-cause format. Specifically, students were able to generate crosses of fruit flies and could observe the effect of the offspring produced, but they did not know the genetic cause. However, prior to the unit students had been taught about genes, DNA, and chromosomes. Therefore, students had the building blocks to begin theorizing about genetic causes. In addition, students had prior experience

360

ECHEVARRIA

doing their own science projects in which they had to generate hypotheses, design a test, and draw conclusions. Thus, students had the basic knowledge necessary to conduct the investigations during the unit. Although students did possess this basic knowledge, they were still considered novices with respect to genetics. As such, it was expected that they would be likely to evidence similar types of novice behaviors with respect to conducting their tests in an unsystematic manner. It was unclear, though, what types of ideas students might construct to explain their data, because no studies have examined student development of the initial baseline simple dominance model of Mendelian genetics.

Inquiry in Mendelian Genetics
The genetics unit in the present study was implemented using an inquiry-oriented approach. According to the National Science Education Standards (National Research Council, 1996), inquiry is defined as the process of “asking questions, planning and conducting investigations, using appropriate tools and techniques to gather data, and constructing and analyzing alternative explanations” (p. 105). All of these processes were incorporated into the unit on genetics. However, an additional aspect of inquiry to consider was the extent to which the phenomena to be investigated could stimulate student curiosity. When students conducted investigations in the present study, some of the inheritance patterns that they encountered appeared predictable on the basis of observable characteristics. However, others were not. It was these unexpected patterns in particular that were used to provoke student curiosity and subsequent explanations. These patterns were considered as anomalous events that were intended to stimulate cognitive disequilibrium in the students (Piaget, 1963). Anomalies in the data were defined as those outcomes that were not readily explainable on the basis of the appearance of the parents. Patterns in which the parents produced offspring with the same appearance were considered “standard” patterns because they were predictable on the basis of appearance (e.g., red-eyed parents produced red-eyed offspring). Anomalous and standard patterns based on parent appearance are shown in Table 1. Lowercase letters denote genetic forms, two per fruit fly. Uppercase letters denote corresponding physical appearance. Notice in Table 1 that red-eyed and white-eyed parents can produce either red-eyed and white-eyed offspring or only red-eyed offspring. This difference in outcomes depends on the genetic makeup of the parents, shown in the left-most column. The apTable 1 Standard and Anomalous Inheritance Patterns for the Mating of Red-Eyed (R) and White-Eyed (W) Fruit Flies
Parent genetic makeup (w-w) (w-w) (r-r) (r-r) (r-r) (r-w) (r-w) (w-w) (r-r) (w-w) (r-w) (r-w) Parent appearance W R R R R R W R R W W R Offspring appearance All W All R All R R, W All R R, W Outcome Standard Standard Standard Standard Anomalous Anomalous

pearance of the fruit fly is dictated by the mechanism of simple dominance Mendelian genetics, in which one genetic form of a trait is dominant to the other. Each fruit fly carries two genetic forms (e.g., red–white or r-w) that make up one gene. The two forms together determine the appearance of the trait. If the dominant form is paired with the recessive form, then the organism will have the appearance of the dominant form. In Table 1, the red form is dominant to the white form. A fruit fly will have the appearance of red eyes if it has a red–red or red–white genetic combination. The fruit fly will only have the appearance of white eyes if it has a white–white genetic combination. It was this structure of dominance and recessiveness that students were intended to theorize about during the unit.

Social Constructivism
During the unit, students were hypothesized to reason and think about ideas based on a social constructivist theoretical framework. Using this framework, construction of meaning was theorized to occur reflexively between the level of the individual and the whole class. In this case, individuals could construct ideas that contributed to whole-class discussions, and the ensuing discussion could contribute to individual ideas (Cobb & Yackel, 1995). Students could also construct meaning in pairs as they worked together on the genetics simulation software. The focus during whole-class discussion was on knowledge construction, specifically on constructing explanations for anomalous data patterns. Students reported on specific data patterns they had encountered and then theorized about potential explanations with the help of the teacher. Students were not provided with formal instruction in Mendelian genetics. Therefore, their explanations were their own constructions. The role of the teachers during the unit was to provide guidance and support for students as they conducted their investigations. This guidance and support generally took the form of suggestions for how to approach the investigations and prompts to think about explanations for the anomalous data patterns encountered. The specific research questions of interest in the study were the following: (a) How will students’ knowledge change in response to anomalous data? (b) How will students reason scientifically in response to anomalies? (c) How are the resulting knowledge construction and scientific reasoning related together?

Method Participants
This study took place in two 7th-grade science classrooms in a large suburban middle school in upstate New York. Each class consisted of approximately 22 to 24 students, ranging in age from 11 to 13 years. The student population of the middle school was predominantly White and middle class. This particular school was selected because the science director is an advocate of constructivist-oriented approaches to teaching science. The science teacher who participated in the study was recommended by the science director as an excellent candidate given the nature of this study. She had taught science at the middle school level for 11 years. This teacher had conducted short inquiry tasks with her students and thus was familiar with allowing students to come up with their own answers to investigations. However, the format of her class was more teacher oriented than the

Note. Lowercase letters denote genetic forms; uppercase letters denote corresponding physical appearance.

KNOWLEDGE CONSTRUCTION AND SCIENTIFIC REASONING format used during the genetics inquiry unit. Thus, the unit represented a fairly new approach to teaching science for both her and her students. Given that this was a new unit for the teacher and her students, I became a participant– observer in the classroom to facilitate implementation of the unit as well as to record data about how the unit was progressing. I have an undergraduate degree in electrical engineering, which included study in several science courses. I am also a university professor of educational psychology specializing in the teaching and learning of science. As a coteacher in the classroom, I was involved with the students on a daily basis and interacted with them one on one during their investigations. I also facilitated whole-class discussions on students’ evolving conceptions of how traits were transmitted in the fruit flies and the plants. As an observer in the classroom, I set up audio- and videotaping equipment, recorded observational notes, and administered pre- and postassessments to the whole class.

361

Materials
To investigate trait transmission in genetics, students used a genetics simulation software, Genetics Construction Kit, and observed the growth of Wisconsin Fast Plants. In both the software and the plants, students investigated or observed the transmission of one trait (e.g., eye color in fruit flies, stem color in plants). The trait could vary with the fruit flies (e.g., wing shape, antennae shape, eye shape) but was fixed with the plants. The plants were included in the unit to provide a living example of an organism that inherited traits based on a simple dominance Mendelian inheritance pattern. However, students could not perform tests with the plants but only observe them. Genetics Construction Kit (GCK). The GCK is a genetics simulation software that students could use to breed fruit flies to observe how traits were transmitted to the offspring. The GCK simulates work done in a genetics laboratory by providing students with an initial “vial” of fruit flies, from which they can produce subsequent vials of offspring by cross breeding two fruit flies. For the purposes of this study, the software was set up so that students could examine simple dominance inheritance patterns for fruit flies that varied on only one trait (e.g., eye color). Figure 1 contains an initial fruit fly vial that students could have seen when they first entered the program. The vial in Figure 1 represents a vial full of fruit flies (symbolized as male and female) that differ on eye color. The first two rows of fruit flies have plum-colored eyes. The second two rows of fruit flies have cardinalcolored eyes. This vial represents an initial field population of fruit flies. This field population is analogous to a scientist gathering a sample of fruit flies, about which nothing is known regarding which parents produced

which offspring. To obtain information about the exact number of fruit flies contained in the vial, students could access a summary chart of each vial. In this case, the chart summarized that Vial 1 contained 25 plum-eyed fruit flies and 8 cardinal-eyed fruit flies. To “breed” the fruit flies, students could select a male and a female from Vial 1 and “mate” them to produce offspring in a second vial. The second vial in Figure 2 contains the offspring of two plum-eyed parents from Vial 1. In this case, both plum-eyed and cardinal-eyed offspring were produced, although once again a majority of the offspring had plumcolored eyes. The information at the bottom of the window indicates which parents were crossed to produce the offspring in Vial 2. In this case, the number 1 followed by a gender symbol and eye color symbolizes that both parents came from Vial 1 and had plum-colored eyes. If students clicked on the gender symbol of one of the parents, they could use the identical fruit fly in a subsequent cross. Using the GCK, students could choose any fruit flies to cross and then analyze the results to see if any particular patterns arose. Thus, they could construct hypotheses and design their own tests to investigate how a particular trait was transmitted. Wisconsin Fast Plants. Concurrent with their investigation of the fruit flies with the GCK, students made observations of the transmission of traits with Wisconsin Fast Plants (Carolina Biological Supply Company, 1989). The plants paralleled the data that the students were gathering from the GCK in that they also illustrated simple dominance inheritance patterns. This parallel was pointed out during whole-class discussion. The Wisconsin Fast Plants varied on one trait, stem color. The stem color could be either purple, which was the dominant variation of the trait, or green, which was the recessive variation. Students observed the growth of three generations of plants. The first generation consisted of purple- and greenstemmed plants. The offspring of those plants consisted of only purplestemmed plants. The third generation of plants consisted again of both purple- and green-stemmed plants. Students collected data on various characteristics of the plants, determining how they were similar and different, and observing how those similarities and differences were transmitted to the next generation.

Procedure
The duration of the unit was 3 weeks, which consisted of fifteen 42-min periods. Preliminary introductory activities for the unit took 2 days and involved a fingerprinting activity and a mental-model-building activity. The fingerprinting activity consisted of analyzing fingerprints on a worksheet to show that individuals are genetically unique, as evidenced by their unique fingerprints. For the mental-model-building activity, a box was given to each dyad with unknown objects inside. Students had to generate hypotheses for what the objects were, indicate how they were testing and gathering information on the objects, and write down the conclusions they were drawing. The next segment of the unit involved students investigating the transmission of traits in fruit flies using the GCK (Jungck & Calley, 1993). As an introduction to the GCK, a video of scientists collecting data on genetics of fruit flies was shown to the students so that they could see what actual fruit flies looked like and could see the vials that scientists used and how they examined the fruit flies (e.g., looking at eyes or wing shape). After the video, the students worked in dyads on the computer to familiarize themselves with the GCK software. Data cards were also handed out so that students could practice recording their crosses, including their hypotheses, tests, and conclusions. The video presentation and initial work with the GCK took approximately 2 days. After working for a day with the GCK, students were provided with handouts of the parts of the fruit flies, a list of fruit fly traits, and further examples of how to fill in the data cards. Students were also given some introductory information and a short presentation on the Wisconsin Fast Plants as well as some ideas for the types of characteristics to observe with the plants. For the next few days, students alternated working on the plants

Figure 1. Initial vial containing field population of plum-eyed and cardinal-eyed fruit flies.

362

ECHEVARRIA For both questions, the trait involved was shape of the fruit fly abdomen. The two variations of abdomen were bobbed and cut. The inheritance patterns were depicted using a three-level tree as shown in Figures 3 and 4. This type of representation had been used during whole-class discussion to depict three generations of fruit flies. The data for these patterns had been generated using the GCK software. Therefore, in addition to the tree diagrams, students were also provided with a printout of the vials produced from the GCK software that corresponded to the tree diagrams. The first question, shown in Figure 3, depicts two bobbed abdomen fruit flies from Vial 1 that were crossed to produce bobbed children (Vial 2). Two bobbed parents from Vial 2 were then crossed to produce bobbed and cut offspring in Vial 4. In contrast, two different bobbed parents from Vial 2 were crossed to produce only bobbed children in Vial 7. For the second question, shown in Figure 4, two bobbed abdomen fruit flies from Vial 2 were crossed to produce bobbed and cut children (Vial 4). Bobbed and cut parents from Vial 4 were then crossed to produce only bobbed offspring in Vial 5. In contrast, a different set of bobbed and cut parents from Vial 4 were crossed to produce both bobbed and cut offspring in Vial 6. This assessment provided information on students’ content knowledge as it related to explaining anomalous and standard patterns, which were the focus during the unit. Incorporated into the first two diagrams on each page was the same Mendelian inheritance pattern tested at the beginning of the unit with plants. Although the content on the postunit assessment differed from that of the preunit assessment, the inheritance patterns that students were asked to explain were parallel to those on the preunit assessment. Different content (i.e., fruit flies instead of plants) was chosen for the postunit assessment for two reasons: (a) to avoid a practice effect from preunit to postunit and (b) to provide a match to the content of the unit given that students were conducting their investigations with fruit flies. Classroom discourse. During the unit, classroom discourse was captured on audiotape. Two dyads per class were audiotaped as they worked on their investigations. These audiotapes also captured general announcements that the teacher and I made to students during the unit. In addition, two episodes of whole-class discussion were either videotaped or audiotaped. Excerpts from these audiotapes serve to illustrate the nature of classroom interactions as they related to both knowledge construction and scientific reasoning. Student scientific reasoning. To gain an understanding of how students were reasoning during the inquiry unit, student data cards and computer logs on diskette were collected. The artifacts provided an indication of student hypotheses, the tests they ran, and the conclusions that they drew. Although the overall sample of students consisted of 22 dyads, data cards were obtained for only 10 dyads, whereas diskettes were obtained for 20 dyads. The discrepancy in data collection arose because students had taken data cards home to use when writing up their reports for the unit and had not brought them back in. Although the remaining subset of data cards represented slightly less than half of the overall sample, it did provide examples of a wide range of student work. Student hypotheses on the data cards varied from no hypotheses to somewhat complex hypotheses. The number of tests produced per dyad, based on the data cards alone, also

Figure 2. Second vial containing plum-eyed and cardinal-eyed offspring from two plum-eyed parents.

and the GCK. Students observed and recorded characteristics, such as height, number of leaves, buds, and similarities and differences between generations with the plants, and recorded their hypotheses, tests, and conclusion for the GCK data. On two separate days, I initiated whole-class discussion of current student findings. The intention during these discussions was to (a) have students describe the results they were getting and (b) generate ideas to explain why those results were occurring. To facilitate these discussions, I prompted and extended student ideas by restating, pointing out anomalous instances of data, and probing ideas for inconsistencies between explanations from one scenario to another. These discussions generally lasted for half a period. During the other half of the period, students either made bee sticks to use when pollinating the plants or listened to a presentation on the parts of the flower and the parts of the bee. Midway through the unit, the regular science teacher presented information on meiosis, which is the process of cell division, whereby cells with a complete set of chromosomes produce sperm and egg cells with half of the number of chromosomes. Students then completed a sticker activity in which they simulated the phases of meiosis. For the last day, students continued gathering data on the plants and the GCK and then took their data home to write up a report of their findings.

Data Collection
This study incorporated a mixed-method, participant– observer design, in which qualitative and quantitative data were gathered during the unit. Data were collected in a pretest–posttest design as well as through artifacts and student discourse during the unit. Content knowledge pre- and postunit measures. To assess students’ content knowledge before the unit, a pre-unit assessment was administered to students during class 1 week prior to the beginning of the unit. The assessment depicted a three-level tree diagram in which a tall and a dwarf plant were crossed to produce only tall plants. The tall plants were subsequently crossed and produced both tall and dwarf plants. The inheritance pattern depicted in this assessment was a standard Mendelian inheritance pattern, with tall as the dominant variation of the trait. Students were asked to explain why the children plants were all tall but the grandchildren plants were both tall and dwarf. Given that the students were to study exactly this type of pattern during the unit, it was relevant to determine whether they could explain this pattern before the unit began. Students took 5–10 min to complete this assessment. A postunit explanation assessment was administered during class the week after the unit ended. Students took approximately 15–20 min to complete this assessment. This explanation assessment was also a paperand-pencil assessment and consisted of two questions in which students were asked to explain several contrasting inheritance patterns for fruit flies.

Figure 3. Tree diagram on postunit assessment depicting different outcomes for the cross of two bobbed fruit flies.

KNOWLEDGE CONSTRUCTION AND SCIENTIFIC REASONING

363

Preunit Explanation Assessment
For the preunit explanation assessment, students were asked to explain why the offspring of a tall and a dwarf plant were all tall plants and subsequently why the offspring of the tall plants were either tall or dwarf. Student responses varied in the extent to which they addressed this question. Some students seemed only to respond to why the children plants were tall; other students only responded to why the grandchildren plants were tall and dwarf, and yet others responded to both situations. Therefore, the overall response for each student was coded and, where possible, distinctions were made in the categories as to whether students addressed both situations or only one in their responses. Interrater reliability of these categories on a subset of 15% of student responses was 0.88. Four explanation categories were formed, as depicted in Table 2. One of the largest response categories (29%) was the overpowering-theory category, which had four subcategories. The overarching idea for this category was that of one gene overpowering another. Underlying this idea was the notion that the children plants received characteristics from both parents but that the characteristic from one parent overpowered the other. This idea was used as an explanation for why the children were all tall. Although all four subcategories encompassed this overpowering notion, only two of the four subcategories also contained an explanation for why the grandchildren then were both tall and dwarf; these are the first two categories listed in Table 2. In the dominant-versusinherited subcategory, students explained that tall was dominant but that the dwarf gene had still been inherited by the grandchildren plants. Therefore, some plants were tall and some plants were dwarf. In the stronger-versus-weaker-strength subcategory, tall was stronger in the child generation but then became weaker relative to dwarf in the grandchild generation; therefore, dwarf plants were able to grow. Responses that referred only to why the

Figure 4. Tree diagram on postunit assessment depicting different outcomes for the cross of a bobbed and a cut fruit fly.

varied tremendously from 9 to 51, with an average number produced of 22.1. This number was slightly higher than the average number of tests produced for the overall sample of 22 dyads, which was 21.0; this number was compiled from saved computer logs in addition to the data cards. Given these parameters, I concluded that this subset was representative of the larger sample. Thus, these data cards were considered a valid subset from which to draw conclusions regarding typical student behavior.

Results
The questions of interest in this study dealt with changes in content knowledge in response to anomalies, approaches to scientific reasoning in response to anomalies, and the relation between the knowledge constructed and scientific reasoning.

Changes in Content Knowledge
The content knowledge that students constructed during the unit was analyzed on the basis of their preunit ideas about trait transmission, homework responses, ideas constructed during wholeclass discussion, and their postunit ideas of trait transmission. Student ideas underwent changes toward greater explanatory power of anomalous data patterns.

Table 2 Explanations for Why a Dwarf and a Tall Plant Would Produce Only Tall Children but Both Tall and Dwarf Grandchildren
Category Overpowering theory Subcategory Dominant versus inherited Stronger versus weaker strength More genes Dominant–stronger Single characteristic Same makes same Genes lost Other pollen Skips a generation Environment Note. n % (No.) 4 (2) 7 (3) 9 (4) 9 (4) 13 (6) 4 (2) 7 (3) 11 (5) 4 (2) Example response S37: “The taller plant gene was more domienent [sic] than the smaller one. But the children still had the smaller plant gene and passed it down to their children and some of them were more domienent [sic] than others.” S14: “I think that all of the children plants grew tall because the tall parent plant has stronger chromosomes. I think the granchildren [sic] were more even because the chromosomes got weaker.” S9: “I think that the children plants grew taller because there were more tall genes.” S29: “I think that the thing that makes a plant grow tall, is stronger than the thing that makes a dwarf plant grow.” S6: “Because the tall ones made the tall ones tall and the opposite for the short ones.” S12: “I think the grandchildren’s heights varied because some of the small parents chromosomes got lost in mitosis, therefore losing the origanal [sic] parents appearance.” S35: “Some of the pollen may have gotten crossed with other pollen that could have affected the growth.” S16: “I think that this happened because it skipped a generation.” S33: “Because the generation of plants may have changed and it might have something to do with the amount of sunlight on each plant.”

45. An additional 14 responses (31%) were coded as don’t know or unintelligible. S# refers to the student number.

364

ECHEVARRIA

child plants were tall suggested that perhaps there were “more tall genes” compared with the number of short genes, thereby causing more tall plants to grow. Other students suggested that tall was simply “stronger” or “more dominant” than dwarf, which caused more tall plants to grow. The second largest explanatory category was the singlecharacteristic category (24%). Underlying this category was the idea that the characteristics from only one parent influenced the appearance of the child, rather than both parents contributing. In the same-makes-same subcategory, students theorized that tall plants would make other tall plants, and dwarf plants would make other dwarf plants. Therefore, the children plants were tall because they only received genes from the tall parent. The grandchildren were both tall and short because some of the grandchildren plants got tall genes but others got dwarf genes. Also in the singlecharacteristic category, some students appeared to try to reconcile the idea that both parents contributed to the offspring even though the offspring only resembled one parent or the other, by suggesting that perhaps “genes were lost.” In this case, although both parents may have contributed genes to the child plant, genes from one parent or another might have gotten lost somehow, resulting in a plant that had the appearance of one parent or the other. Related to this idea was the other-pollen category, where instead of referring to genes, students referred to the plant pollen in a similar sense. In this case, students again seemed to acknowledge the idea that both parent plants contributed to the appearance of the children plants but indicated that perhaps the pollen of one or the other parent plant was not effective or got mixed up with other pollen to cause unpredictable results. A small category that was more descriptive than explanatory was the “skips a generation” (11%) response where students suggested that the results came out as they did because traits may skip a generation. This response is valid because it does describe what happened with the plants, in that the dwarf characteristic skipped the child generation but reappeared in the grandchild generation. However, it does not explain how that might have happened. Another small category was coded as environment (4%) because students appealed to environmental factors that might have caused differential growth rates to explain the disparity in heights between the children and grandchildren plants. The responses in this category seemed to indicate a certain amount of confusion with respect to the idea that a dwarf plant was short because of a gene that caused it to grow short, rather than because it lacked sunlight or because it was not full grown. This type of thinking seemed to be fairly concrete and was in marked contrast to the responses of the other students. The remaining responses were coded as don’t know– unintelligible (31%) because students indicated that they did not know why all the children plants grew tall and the grandchildren plants grew both tall and small, or the responses were unintelligible. Of the ideas presented on the explanation assessment, students who proposed some sort of overpowering theory presented the most coherent ideas in that their theories could explain the pattern of data while still retaining the canonically correct notion that offspring received genetic contributions from both parents. This notion was not retained by the students who proposed a samemakes-same idea where only one parent would contribute their genes to the offspring, although this theory would also explain the data. The genes-lost and other-pollen categories presented a mix of

ideas where both parents contributed but somehow the contribution of only one parent was effective, rather than both. For those students who proposed that perhaps the trait skips a generation, it was not possible to determine by what mechanism they thought that this might occur. Responses coded in the environment category seemed out of sync with the rest of the class with respect to the concreteness of the answers, although they did make sense on the basis of concrete logic. That is, lack of sun and water would cause a plant to grow in a distorted manner, perhaps making it much smaller than it would otherwise normally be. However, genetic causes and not environmental causes were being sought on this assessment.

Conceptions During the Unit
During the unit, students worked with the genetics software to gather data on fruit-fly crosses and made observations of the growth of the plants. The goal of the unit was for students to construct mental models to explain the patterns of data that they encountered. The roles of the teacher and I during this time were to provide guidance and support to the students as they were working on their investigations. Our main objective was to help them focus on patterns in the data and to help them develop explanations for those patterns. To do so, we used a homework assignment, made suggestions about developing explanations, and used whole-class discussion to further support them in explanation building. While the students worked in dyads, we also circulated throughout the room answering questions and offering feedback or suggestions. Homework patterns. For the homework assignment, students were asked to describe any patterns that they had seen in their data. This assignment was open ended in nature. As a result, students were free to write as little or as much as they chose on any particular aspect of the data they had gathered. To analyze these data, each type of pattern that students described was coded and tallied to get a frequency count on the number of students noticing that pattern. Each student’s response generally included the description of several different patterns and so included several different codes. The codes were sorted into four main categories, each with subcategories. These results are reported in Table 3, where each student’s response could have contributed to more than one category. The first category was coded as standard inheritance patterns. This category consisted of inheritance patterns in which offspring were produced by parents that looked like them. Sixty-one percent of the student responses described the standard pattern of crossing one type of parent to produce one type of offspring. Forty-four percent of the responses described crossing two types of parents to produce two types of offspring. The second category was coded as anomalous inheritance patterns. This category consisted of inheritance patterns in which offspring were produced that did not look like the parents or that looked like only one of the parents. Thirty-seven percent of the responses described a situation in which one type of parent produced two types of offspring. Thirty-two percent described a situation in which two different types of parents produced only one type of offspring. Students also noticed differences in the relative number of the two types of offspring produced. Twenty percent of student re-

KNOWLEDGE CONSTRUCTION AND SCIENTIFIC REASONING

365

Table 3 Student Homework Responses Describing Inheritance Patterns Seen During Their Investigations
Inheritance pattern category Standard patterns One type produces one type Two types produce two types Anomalous patterns Two types produce one type One type produces two types Number patterns More of one type of offspring than another When crossing different types, even amount of each type in children Gender patterns: Male or female type more prevalent Note. n 41. % (No.) 61 (25) 44 (18) 37 (15) 32 (13) 20 (8) 12 (5) 12 (5) Example response S4: “We crossed rosy with rosy and got all rosy.” S21: “Then we crossed Threads and Aristaless and we got both.” S27: “Then we crossed 2 Kidneys from the last experiment, and got both [Kidney and Dach] again.” S21: “We crossed Aristless [sic] and thread again and we got only thread.” S13: “There is always one type of fruitfly that there is more of (or just that type of fruitfly) in each cross.” S7: “When we crossed a Aristopedia and Thread we would get a close to even amount of each.” S18: “We have seen that when you cross the flies there’s more of the female type.”

Each student response could be coded in more than one category. S# refers to the student number.

sponses indicated that one type of offspring was produced more than another. This was an especially relevant pattern to notice in order to generate the idea that both variations of the trait did not always behave in the same way. Students also noticed that two types of offspring could be produced in approximately equal amounts when crossing two different types of parents. Both of these patterns have a canonical basis, in that the dominant appearance will tend to be more prevalent than the recessive appearance when offspring are produced, but can also occur in equal amounts to the recessive appearance depending on the genetic makeup of the parents. A pattern that relatively few students commented on was that the appearance of the traits seemed to vary on the basis of gender. Responses in this category described outcomes in which more of the female’s type or more or the male’s type was produced. Although in general there are some traits that are “sex linked,” meaning that they occur with a higher frequency in either males or females, the traits examined in this context were not sex linked. Thus, although some students noticed a covariation of traits with gender, this would not have been a fruitful avenue to continue exploring because it was not causally related to the patterns of inheritance. With the exception of the gender patterns, all other patterns that students noticed were relevant with respect to simple dominance trait transmission. Teacher suggestions and dyad interactions. In addition to focusing on patterns, the teacher (LG) and I (ME) would typically make announcements to the class as a whole reminding them to think about explanations for their data. The emphasis on knowledge construction was specifically on generating explanations for the outcomes that the students observed and testing those explanations. For example, when the students began working with the fruit flies, I suggested how they might approach their investigations when they were beginning to test a new trait.
When you start with a new trait . . . you’re going to have a certain theory about how traits are transmitted based on the first trait that you worked with. When you go to the second trait, see if the same kinds of things happen. See if you can test out your theory and predict what’s going to happen. (ME)

your ideas about your results and patterns . . . . Can you explain these patterns. Brainstorm with each other” (LG). Suggestions such as these continued throughout the unit, with both the teacher and I emphasizing that the objective was for students to collect data and come up with explanations for the patterns that they saw. Although the teacher and I were encouraging students to theorize with each other, this seemed to occur to varying degrees depending on the dyad. In one of the audiotaped dyads, the students theorized that gender played a role in whether one form of the trait appeared more than the other. Specifically, one of the students suggested to her partner, “I kind of think that females, well females lay eggs, so it might have more” (A). However, in another audiotaped dyad, student explanations were not overt in their dialogue. Rather, these students seemed focused on making predictions that were either correct or incorrect, noted the outcome as such, and moved on. In the following excerpt, the students (M and R) crossed a cut-winged fruit fly with a wrinkled-winged fruit fly. They predicted that there would be more cut-winged than wrinkled-winged offspring:
M: So what do you think? More wrinkled or more cut? R: Um, so far there’s been more cut.

M: Yeah, there’s always. So more cut? R: Okay, cross two.

M: They’re all cut. Okay, analysis. R: We were sort of right?

M: Sort of right. R: Our hypothesis was not exactly right. We had no wrinkled. That’s a lot of cut.

M: You want to go on to something else now?

During the unit, students were also encouraged to talk with each other and generate explanations. “Work with you partner. Share

Although the students predicted more cut-winged than wrinkledwinged offspring, they found that the cross produced only cutwinged fruit flies. This outcome illustrated the dominant anomalous outcome in which two different types of fruit flies (i.e., cut,

366

ECHEVARRIA

wrinkled) produced only one type of offspring (i.e., cut). The students acknowledged that this was not the outcome they were looking for (“We were sort of right”), but they moved on nevertheless. In this case, it was not possible to determine what explanations they may or may not have been forming regarding this particular outcome. Whole-class discussion. To aid students in generating explanations for the outcomes they observed, I used whole-class discussions. My role in these discussions was as facilitator. I first solicited student comments on particular patterns they had observed and wrote these patterns on the board. I then questioned them about explanations for the anomalous instances in particular. In the following scenario, a student mentioned that she had crossed two bobbed-abdomen (B) fruit flies to produce both bobbed- and narrow-abdomen (N) fruit flies. This cross was written up on the board as B B 3 B, N. When I prompted students for an explanation of this pattern, the following responses were offered by Amelia (A) and Brenda (B) in response to the question of why two bobbed-abdomen fruit flies produced offspring that had both bobbed and narrow abdomens:
A: Well, when we took all the fruit flies from the field population, maybe the two that you crossed, maybe one of the fruit fly’s parents was a different type, so that its genes will be different, so that it produces some of the other type of species.

Here students referred to a stronger type of the gene, which is analogous to the dominant variation. They further implied that there was also a weaker type, which is analogous to the recessive type. Thus when a stronger type and a weaker type were paired, the fruit fly would have the appearance of the stronger type. These ideas were all approximations of canonically correct notions.

Postunit Explanation Assessment
At the end of the unit, students were administered a postunit explanation assessment to determine whether they had made gains in their ability to explain the anomalous data patterns. For this assessment, students were asked to explain the outcomes for contrasting standard and anomalous inheritance patterns. Given the focus on anomalous data patterns, only data pertaining to explanations for the anomalous inheritance patterns are discussed. The anomalous inheritance patterns that students were asked to explain were parallel to the preunit patterns, but instead of plants, students were asked about fruit flies and abdomen shape. For the first question, students were asked to explain why two bobbedabdomen fruit flies produced both bobbed- and cut-abdomen offspring. For the second question, students were asked to explain why a bobbed- and a cut-abdomen fruit fly produced only bobbedabdomen offspring. Student responses for each question were coded and categorized. Interrater reliability on the resulting categories for a stratified sample of 15% of the responses was 0.93. Responses to the first question were sorted into two main explanatory categories and are depicted in Table 4. The first category was the mixed-breed category. Responses were coded in this category if they indicated that the reason two bobbed-abdomen fruit flies could produce both bobbed and cut offspring was because one or both of the parents carried the cut variation. This explanation closely parallels the canonical explanation, in which both parents must carry the cut variation in order for both bobbedand cut-abdomen fruit flies to be produced. Fully 78% of student responses were coded in this category. However, the mechanisms underlying the explanations varied. In most cases, students simply indicated that the cut variation was present in the parents even though the parents had the appearance of bobbed-abdomen fruit flies. A few responses suggested that all of the fruit flies carried both variations and that it was a matter of chance which variation became apparent in the children. Finally, some students also suggested that both parents carried the cut variation and that it was an accumulation of the cut variation that caused cut offspring to be produced. The remaining 22% of the responses were coded according to various nongenetic or unintelligible categories. For the second question, students were asked to explain why a bobbed and a cut fruit fly produced only bobbed offspring. Responses to this question were sorted into three main categories, as summarized in Table 5. Student responses in the dominant category indicated that bobbed was dominant as an explanation for why only bobbed offspring were produced. This response approximates the canonical explanation, in that the reason that there are only bobbed offspring is that the offspring all carry a bobbed– cut genetic combination but appear as bobbed because bobbed is dominant to cut. The student notion of dominance, however, was manifested in various ways. Some students simply stated that bobbed was dominant in the offspring. Other students described

ME: So are you’re saying that these [pointing to bobbed parents on chalkboard] might not be identical? What’s going on with this bobbed? A: That would have to be, it’s like, if you think of it, what chemicals it has, and say it has a little bit of Chemical N in it.

ME: So it might have some narrow, some narrow might be in this [in one of the bobbed parents]? Okay, other ideas? Brenda? B: It’s like, that bobbed carried narrow, it didn’t get that, but it still has that as part of its genes so it just carried through.

These explanations were significant because they were alluding to the idea that a fruit fly could carry both genetic variations of the trait but only look like one variation. That is, a fruit fly could look like a bobbed fruit fly but still carry the narrow genetic variation. This idea was canonically correct. A fruit fly that has a bobbed and narrow variation, where bobbed is dominant to narrow, will have the appearance of a bobbed fruit fly but will also carry the narrow variation that can be passed on to the offspring. To follow up with this idea, I then asked the students why a bobbed fruit fly looks as it does if it has some narrow in it. Amelia (A) and Carrie (C) offered responses:
ME: Okay, why is it then, if this has some narrow in it, why does it look like a bobbed? Amelia? A: Well, because the bobbed, the B or whatever, probably has stronger chemical genes in it so its appearance look like it’s B, but it does have some of the N.

ME: Okay, so somehow it’s stronger? Carrie? C: Well, like I said, like one of the parents is probably bobbed, and it probably got like, like it is probably a stronger trait, and it probably got more of the bobbed trait in it and probably some of the trait from the narrow.

KNOWLEDGE CONSTRUCTION AND SCIENTIFIC REASONING

367

Table 4 Student Responses to Why Two Bobbed Fruit Flies Produced Both Bobbed and Cut Offspring
Explanation Mixed breed % (No.) 78 (33) Example response S1: “The reason for this is probly [sic] because the parents of one had cut geens [sic] in it.” S46: “I think the reason is chance by this I mean all the flies have 2 different traits in them and it’s a matter of chance witch [sic] one comes up.” S37: “Both the parents and children might have some of the cut gene but it wasn’t strong enough to form a child [in Vial 2] but in the grandchild all of that formed together to make a child [Vial 4].” S10: “Vial 4 has more types than Vial 7. Vial 7 has no cut.” S18: “The bobbed was a female mabie [sic] in Vial 2.” S23: “This happened because the dominate [sic] gene was more effective to the bobbed parents even though cut was involved.”

Nongenetic Description Gender Unintelligible Note. n 43.

9 (4) 2 (1) 12 (5)

S# refers to the student number.

dominance as a greater amount (i.e., there were more bobbed genes) or as a greater strength (i.e., bobbed was stronger than cut). For the mixed-breed response, students suggested that either one or both fruit fly parents carried both traits but that the offspring happened to come out bobbed. The mechanism underlying this category seemed to be that of chance. For example, some students indicated that the cut fruit fly also carried the bobbed variation and that the offspring therefore could be bobbed. Alternatively, they suggested that both fruit flies carried both forms of the trait, and it was a matter of chance which form became apparent in the offspring. Therefore, to explain the anomalous outcome, students simply assumed that by chance the offspring all came out bobbed. Comparison of the dominant and mixed-breed categories for this question indicates a trend toward dominance as the preferred explanation for this outcome compared with a mixed-breed explanation (21 vs. 11), 2(1, N 31) 3.13, p .08. Attributing the anomalous outcome to dominance would be the appropriate canonical response. The remaining 21% of the responses were coded according to various nongenetic or unintelligible categories. Change in content knowledge. Responses for the pre- and postunit explanation assessments were examined to determine whether students had made gains in their ability to explain the anomalous inheritance patterns. Only those students who had completed both assessments were included for this analysis. To make a parallel comparison, student responses for the first expla-

nation assessment were compared with the responses for the corresponding anomalous inheritance patterns on the second explanation assessment. Thus, a student’s response explaining why a tall and a dwarf plant produced only tall offspring was compared with the same student’s response explaining why a bobbed and a cut fruit fly produced only bobbed offspring. Likewise, the response to why two tall plants produced both tall and dwarf plants was compared with the response to why two bobbed fruit flies produced both bobbed and cut offspring. To analyze both assessments, the rubrics for the preunit and postunit assessments were combined to form a single rubric. The resulting rubric consisted of descriptive categories that were subsequently ranked from 0 (low) to 4 (high) on the basis of explanatory power and relevance to canonical concepts. Categories are shown in Table 6. Responses that received a ranking of 0 were those that were descriptive rather than explanatory, nongenetic, unintelligible, or in which the student specifically stated that he or she didn’t know how to explain the pattern. Responses that received a ranking of 1 were those that suggested that the offspring received characteristics from only one parent rather than both. This type of response was apparent only on the preunit assessment. Responses that received a ranking of 2 were those that explained either one type of anomalous pattern or another but not both. Responses that received a ranking of 3 were those that explained both patterns using some form of a mixed-breed explanation rather than citing dominance as an explanation. These responses were

Table 5 Student Responses to Why a Bobbed and a Cut Fruit Fly Would Produce Only Bobbed Offspring in Vial 5
Explanation Dominant % (No.) 53 (21) Example response S17: “In Vial 5 bobbed was the dominant gene.” S35: “This may have happened because in Vial 5 there may have been more bobbed genes . . . So it might have depended on the amount of genes in the parents.” S2: The reason for this was because probly the bobbed in [Vial 4] was proby had stronger chinicial geens so all the trates came out to be bobbed [sic].” S26: “In Vial 5 the cut parent had bobbed in it.” S10: “Vial 6 has both types and Vial 5 has only one type. They’re both different.” S18: “Maybe the female was the bobbed in Vial 2 and the cut in Vial 4.” S36: “In Vial 5, the only bobbed were the offspring because the cut relationship was erased because of two sets of bobbed parents.”

Mixed breed Nongenetic Description Gender Unintelligible Note. n 40.

28 (11) 7 (3) 2 (1) 10 (4) S# refers to the student number.

368

ECHEVARRIA

Table 6 Frequency of Student Explanations for Anomalous Inheritance Patterns From Preunit to Postunit
Preunit Score 4 3 2 1 0 Category Dominant–stronger versus mixed breed Stronger versus equal strength Mixed breed–mixed breed Dominant Mixed breed Single characteristic Unintelligible Don’t know Environment Gender Skips a generation–description % (No.) 5 (2) 8 (3) 17 (7) 25 (10) 5 (2) 28 (11) 3 (1) 13 (5) Postunit % (No.) 34 (14) 12 (5) 24 (10) 5 (2) 7 (3) 7 (3) 3 (1) 8 (3)

Scientific Reasoning in Response to Anomalies
The processes of interest with respect to student scientific reasoning were the generation of hypotheses and design of tests. Particular attention in these analyses was paid to the generation of anomalous versus standard hypotheses and design of tests that would produce anomalous versus standard outcomes. Teacher suggestions and dyad interactions. With respect to conducting their investigations, the teacher and I made various suggestions to the students regarding comprehensiveness and consistency of testing with the intention of facilitating discovery and testing of anomalous outcomes. These suggestions were generally prompted by either observations of student progress, questions that students had asked us, or exchanges that we had with students. For example, one suggestion was prompted by an exchange that the teacher had with 2 students who had been crossing fruit flies in a systematic manner. During this conversation, the students shared their testing plan with the teacher in which they were crossing two like forms of the trait (e.g., two red-eyed flies or two white-eyed flies) and then crossing two unlike forms of the trait (e.g., a red-eyed and white-eyed fly). The teacher confirmed for them that this was an excellent approach and then made an announcement to the class describing the students’ plan. She concluded by pointing out that these 2 students had a strategy for their testing and suggested to the class that students should have a plan for their investigations (LG). Systematically crossing all possible combinations of the trait would help students to discover anomalous outcomes. In addition to an overall testing plan, both the teacher and I also suggested to students to try the various types of crosses several times with different fruit flies to determine whether the results they encountered were consistent. This suggestion was also intended to help students discover anomalous outcomes and to see that they occurred with some level of consistency. I also pointed out to students that they could vary one fruit fly at a time in their crosses by selecting the same fruit fly parent from one cross to use in another. My intention with this suggestion was to help students to discover that certain fruit flies could contribute to particular types of outcomes. As students worked on the GCK, the teacher and I circulated throughout the room checking on student progress, answering questions, and offering suggestions. As an example, in the follow-

Note. n 41. The shift from preunit to postunit was significant at p .001 using a Sign Test.

only apparent on the postunit assessment. Responses that received a ranking of 4 were those that most closely approximated canonical explanations for both inheritance patterns, which included citing both dominance and parents that carried the other form of the trait (mixed breed) as rationales to explain the anomalous inheritance patterns. An alternative explanation that also approximated the canonical concept was the stronger-versus-equalstrength response in which students suggested that bobbed and cut fruit flies only produced bobbed offspring because bobbed was stronger than cut. In contrast, two bobbed parents produced bobbed and cut offspring because bobbed was relatively weaker in that situation, which would “allow” both bobbed and cut offspring to be produced. The change from preunit to postunit rankings is graphed in Figure 5. A Sign Test was used to analyze the ranked data. The results of this test indicated that student responses showed significantly more explanatory power on the postunit assessment compared with the preunit assessment with respect to explaining the anomalous genetic patterns (Sign Test 28, p .001). Asymmetric development of explanations for anomalous patterns. Although students apparently made a significant shift toward more explanatory power of anomalous outcomes, examination of their explanations indicated that they did not do so in a symmetrical manner. Specifically, on the postunit explanation assessment, 78% of the students were able to posit a mixed-breed explanation for the occurrence of two bobbed fruit flies producing bobbed and cut offspring (see Table 4). In contrast, 53% of the students were able to posit a dominant explanation for the occurrence of a bobbed and a cut fruit fly producing only bobbed offspring. Another 28% of the students posited a mixed-breed explanation for this same outcome, rather than positing dominance (see Table 5). Why were 78% of the students able to construct the mixed-breed explanation, whereas only 53% of the students were able to construct the dominance explanation? Student patterns of scientific reasoning were examined next in an effort to shed light on this question.

Figure 5. Frequency graph of pre- and postexplanation assessment scores. 0 low and 4 high.

KNOWLEDGE CONSTRUCTION AND SCIENTIFIC REASONING

369

ing exchange students had come across a mixed-breed anomalous outcome. I noticed this outcome as I walked by and made a suggestion for how they might further investigate it. In this scenario, Lauren and Ricky had crossed two brown-eyed fruit flies; all brown-eyed offspring were produced. They repeated this cross with a different set of fruit flies and obtained the same outcome. They tried the cross again with another set of fruit flies but found that instead of all brown-eyed offspring, both browneyed and purple-eyed offspring were produced. Lauren (L) was puzzled but did not seem sure what to do. Ricky (R) seemed unperturbed by the anomaly.
L: R: L: They’re mixed this time. That’s weird. What vial number did we do it from? From 2 and 3. They’re both brown [-eyed parents] and now we got purple. Analysis chart? Oh, sorry. Eight. That’s really weird, like that they were mixed this time. Eight [recording number of purple-eyed fruit flies]. . . . and 37 [brown-eyed]. Total is 45 . . . . Calculator? This recorder doesn’t have a little red button. Yeah. What percentage? The eight is 17% . . . 17.7%, 82.3% [recording percentage of each type of fruit fly]. [I walk by and notice that the students have come across an anomalous outcome.]

parent gave Lauren the impetus she needed to further examine the unusual outcome she encountered. After I left, Lauren and Ricky decided to cross the same parent from Vial 2 with another browneyed parent. They found that only brown-eyed offspring were produced. When they crossed the parent from Vial 3 with a different brown-eyed parent, they found the same unusual outcome. Therefore, Lauren concluded the following to Ricky:
L: So, I think it’s just that one parent, number three, that makes it, like I think they have like background or something, so it’s just that one parent that makes it do that [produce both brown- and purple-eyed offspring].

R: L:

R: L: R: L: R: L:

ME: So what . . . I’m seeing something interesting here. You’re crossing brown-eyed and brown-eyed and you got both. Any ideas why? What’s going on there? R: Something with the field population because that was mixed. Maybe it skipped a couple of generations?

ME: Okay, now you took it out of Vial 2 and Vial 3? L: Yeah.

ME: Ok, how else might you explore that because you know you can select that exact same fruit fly to cross with another fly . . . [The conversation is interrupted, as the teacher tells students to get out notebooks to write down homework questions.] [I move on.] L: Alright. I think what we should do is use that same parent, like use the parent from Vial 2, the parent that we used from Vial 2 again with some other one to see if it’s just that parent, you know, and then we can try it with the one from Vial 3. It could have something to do with just that parent, you know?

Although my exchange with Lauren and Ricky was interrupted, my suggestion to cross one of the same parents with a different

In this particular scenario, what Lauren determined was that there was a type of fruit fly that seemed to be able to produce both types of offspring. She hypothesized that this particular fruit fly had “background or something” that contributed to this outcome. That she arrived at this conclusion illustrates the dynamic between instruction and dyadic inquiry as well as between knowledge construction and scientific reasoning. Specifically, students were encouraged to comprehensively and consistently test the combinations of each trait. She and Ricky did precisely that and discovered an anomalous outcome. Given the emphasis on anomalies as a catalyst for knowledge construction and scientific reasoning, I tended to circulate and specifically look for such occurrences as a means of prompting further thinking on the part of the students. My intervention when I observed the anomalous outcome helped these students to further investigate the anomaly and subsequently develop the idea that a fruit fly with a particular appearance could produce offspring with a different appearance. Pattern of anomalous and standard hypotheses. Given the emphasis on investigation of anomalous outcomes, I was interested in determining whether, and to what extent, student hypotheses included predictions of anomalous outcomes. To analyze student hypotheses, categories were formed based on the data cards generated per dyad. The categories corresponded to the standard and anomalous inheritance patterns found in Mendelian genetics. Two types of standard hypotheses were coded. One hypothesis involved predicting that parents of the same type would produce only that type of offspring (e.g., red-eyed parents will produce only red-eyed children). The second hypothesis involved predicting that two parents of different types would produce offspring of both types (e.g., red- and white-eyed parents will produce both red- and white-eyed children). Anomalous hypotheses were those in which it was predicted that offspring would be produced that did not look like the parents or that looked like only one of the parents. Two types of anomalous hypotheses were coded. One hypothesis involved predicting that two parents of different types would produce only one type of offspring (e.g., red-eyed and white-eyed parents will produce only red-eyed children). This anomalous hypothesis is labeled with the analytic term of dominant hypothesis because it is accounted for by dominance. The second hypothesis involved predicting that parents of the same type would produce offspring of different types (e.g., two red-eyed parents will produce both red-eyed and white-eyed children). This anomalous hypothesis is labeled with the analytic term of mixed-breed hypothesis because it is accounted for by the fact that the parent fruit flies carry both genetic forms of the trait.

370

ECHEVARRIA

For each dyad, the total number of anomalous and standard hypotheses were compiled. Average totals are in Table 7. Using a Friedman analysis of variance (ANOVA), there was a difference in frequency with which students proposed the different types of hypotheses, 2(3, N 9) 10.0, p .02. Using a follow-up Wilcoxon signed-ranks test, both types of standard hypotheses were produced significantly more frequently than the dominant hypotheses (Z 2.52, p .01). The mixed-breed hypotheses were also produced significantly more frequently than the dominant hypotheses (Z 2.04, p .04). However, the standard hypotheses were not produced significantly more frequently than the mixedbreed hypotheses. Of the two anomalous hypotheses, only the dominant hypothesis was produced less frequently than the standard hypotheses. The mixed-breed hypothesis was produced at a comparable level with the standard hypotheses. Thus, the production of hypotheses mirrored the asymmetry found on the postunit explanation assessment. Specifically, students proposed more mixed-breed than dominant explanations on the postunit assessment. Likewise, students generated more mixed-breed than dominant hypotheses during their investigations. Testing of anomalous and standard outcomes. To further investigate the pattern of hypotheses generated by the students, testing patterns were then examined to determine the types of tests that students elected to run. Three categories were formed from the computer logs and data cards of the tests that students had run. These three categories corresponded to the different possible combinations of tests that students could run of one trait with two different variations (e.g., R R, R W, W W). Particular attention was paid to the tests that could produce anomalous outcomes. This breakdown was of interest because of the premise that anomalous outcomes were most likely to provoke cognitive conflict. Therefore, it was of interest to note whether there was differential testing of the combinations that would tend to produce anomalous outcomes compared with the combination that was not likely to produce an anomalous outcome. The types of tests that could produce anomalous outcomes were those of the dominant form crossed with itself or the dominant form crossed with the recessive form (e.g., R R or R W). The recessive form crossed with itself would always produce itself and therefore was not considered anomalous.

For each dyad, the dominant and recessive variations of each trait tested were determined on the basis of the tests and outcomes that had been generated for that trait. The number of Dominant Dominant, Dominant Recessive, and Recessive Recessive tests were then recorded for each trait tested per dyad. The average number of each type of test run per dyad is recorded in Table 8. Using a Friedman ANOVA, there were significant differences between the frequencies with which each type of test was run, 2 (2, N 22) 7.8, p .02. In follow-up analyses with the Wilcoxon signed-ranks test, the dominant dominant tests were conducted significantly more frequently than the recessive recessive tests, Z(21) 2.9, p .005, but the dominant recessive tests were not conducted significantly more frequently than the recessive recessive tests, Z(21) 1.6, p .10. The dominant– dominant cross is the test that can produce the mixed-breed anomalous outcome. This combination was tested significantly more frequently than the recessive–recessive combination. The dominant–recessive combination is the test that can produce the dominant anomalous outcome. This combination was not tested significantly more frequently than the recessive– recessive combination, although there was a trend in that direction. Thus, students were most frequently conducting the test that produced the mixed-breed anomaly. Next in frequency, students conducted the test that produced the dominant anomaly. Least frequently, they conducted the test that produced no anomalies. Differences in anomalous outcomes. Given the finding that the dominant– dominant tests were run more frequently than the recessive–recessive tests but the dominant–recessive tests were not, an additional analysis was run to examine what types of outcomes students actually saw. The frequency of mixed-breed anomalous outcomes as a proportion of the dominant– dominant tests conducted was calculated for each dyad and compared with the frequency of dominant anomalous outcomes as a proportion of the dominant–recessive tests conducted. The average percentages were 62% and 28% for the mixed-breed and dominant anomalous outcomes, respectively. That is, on average 62% of the dominant– dominant tests conducted resulted in mixed-breed anomalous outcomes, whereas on average 28% of the dominant–recessive tests conducted resulted in dominant anomalous outcomes. Using a Wilcoxon signed-ranks test, the mixed-breed anomalous outcomes

Table 7 Average Number of Anomalous and Standard Hypotheses Generated Per Dyad
Hypothesis type Standard One type produces one type Two types produce two types Anomalous Two types produce one type (dominant hypothesis) One type produces two types (mixed-breed hypothesis) Total Average 6.7a 6.9a 0.4b 3.1a 17.1 .05 using a Wilcoxon signed-ranks test. Example S26/S38: “I think if you mix 2 scute you will get scute children.” S12/S9: “If I cross 1 waxy & 1 heldout the offspring will be both.” S15/S25: “I think when we cross both [types] it will have an offspring of only Aristaless.” S46/S29: “It will come out mixed [when crossing two grooveless].”

Note. n 9 dyads. Different subscripts differ significantly at p S# refers to the student number.

KNOWLEDGE CONSTRUCTION AND SCIENTIFIC REASONING

371

Table 8 Average Number of Each Type of Test Run Per Dyad
Test type Anomalous outcomes possible Dominant Dominant (mixedbreed outcome possible) Dominant Recessive (dominant outcome possible) Standard outcomes only: Recessive Recessive Total tests Average Example test for eye color

7.8a 7.6a,b 5.5b 20.9

Red Red White

Red White White .05

Note. n 22 dyads. Different subscripts differ significantly at p using a Wilcoxon signed-ranks test.

occurred significantly more frequently than the dominant anomalous outcomes, Z(21) 3.1, p .002. Therefore, when conducting a dominant– dominant test, which was the most frequently run test, students were significantly more likely to see an anomalous outcome than when conducting a dominant–recessive test.

a mixed-breed red-eyed fruit fly will be produced. A less sophisticated approach to testing was predominantly to select parents from Vials 1 and 2. These students seemed to be focusing on parent child outcomes rather than on generational outcomes. Lastly, in some cases it appeared that students were selecting parents at random from various vials, on the basis of whichever form of the trait they were interested in crossing. When conducting tests with the genetics software, students were able to perform a cross and then were able to select one of the same parents again to cross with a new parent. In this way, students could perform controlled tests by varying only one parent at a time. This approach would allow them to isolate fruit flies that tended to produce particular results. I had informed students of the possibility of doing this early on in the unit. However, students did not consistently perform controlled tests. Twelve of the 22 dyads used this approach on occasion, but 10 dyads did not use it at all. It is important to note, though, that using a controlled approach was not necessary for students to draw valid conclusions from their data. Rather, it was one approach that could have been used.

Relation Between Knowledge Constructed and Scientific Reasoning
The findings depict an interrelated picture between knowledge construction and scientific reasoning. On the basis of the content knowledge assessments, students constructed explanations for the anomalous outcomes in an asymmetric fashion. More students constructed an explanation for the mixed-breed anomalous outcome compared with the dominant anomalous outcome. Similarly, students generated the mixed-breed hypothesis with greater frequency relative to the dominant hypothesis. They also ran the test that could produce a mixed-breed anomalous outcome significantly more often than the test that would not produce an anomalous outcome, but the test that could produce a dominant anomalous outcome was not run significantly more frequently. Finally, it was found that students saw the mixed-breed anomalous outcome significantly more frequently than the dominant anomalous outcome. Thus, it appears that more students constructed an explanation for the mixed-breed outcome because it was observed significantly more often than the dominant outcome. Further, once students had constructed the notion of a mixed-breed fly to explain the mixedbreed anomalous outcome, they attempted to generalize that explanation to the dominant anomalous outcome, rather than constructing a different explanation for that outcome. This would explain why 28% of the students posited a mixed-breed explanation for the dominant anomalous outcome on the end-of-unit explanation assessment. Therefore, we saw patterns of student testing that were influenced by attention to anomalous outcomes. This attention, in turn, influenced their knowledge construction. This outcome is similar to a bootstrapping process (Schauble, 1996) in which knowledge influences testing approaches, which then provide additional knowledge.

General Testing Approaches
To provide further information about testing approaches, other aspects of student testing practices were examined. At the beginning of the unit, some students did not comprehensively test their traits; during whole-class discussion, several dyads did not have data for all three combinations of test types for the traits they had investigated. Students did remedy this problem, however, with 21 of the 22 dyads comprehensively testing at least one trait by the end of the unit. Testing all combinations of the trait was mentioned by the classroom teacher early on in the unit when she noticed a pair of students specifically using that approach in their testing. I also mentioned it during the second whole-class discussion when we talked about how to analyze the data by grouping them according to the various types of test combinations. When conducting their tests, students used various approaches for the selection of the parent fruit flies. The most relevant testing approach for students to notice generational trends was to select parent fruit flies from the last vial that was created. For example, students might select two parents from Vial 1 to create Vial 2. They then might select two parents from Vial 2 to create Vial 3 and so forth. Using this form of testing, students were most likely to notice the skips-a-generation pattern. The majority of the dyads (17/22) used this testing approach at least some of the time, if not all of the time. An alternative to a generational approach to testing was to select parents on the basis of the makeup of a particular vial. For example, some students indicated in their hypotheses that they were specifically selecting fruit flies from vials that had a particular combination of parents and offspring. These students had either noticed covariation differences in the likelihood of different types of outcomes depending on the makeup of the parent vial, or they were hypothesizing that such differences might occur. This was an astute observation because the chances of selecting a pure breed (e.g., r-r) or mixed-breed (e.g., r-w) fruit fly do differ on the basis of the type of parents crossed. For example, a pure-breed red-eyed fruit fly can never be selected from the offspring of a vial in which the parents are a red-eyed and a white-eyed fruit fly. Only

Discussion
During the unit in genetics, student ideas underwent changes toward greater explanatory power of anomalous data patterns. At the beginning of the genetics unit, a small proportion of students

372

ECHEVARRIA

(11%) was able to provide an explanation for both anomalous data patterns. By the end of the unit, 70% of students were able to provide a logical explanation for both patterns. Of that 70%, 46% were able to provide an explanation approximating a canonical explanation for both data patterns. An additional 24% had constructed a new explanation for the anomalous data patterns that had not been constructed at the beginning of the unit. This explanation consisted of a mixed-breed mixed-breed mental model in which students posited that all fruit flies carried both forms of the trait and that it was a matter of chance which form became apparent (see Table 6). This is a valid and logical notion that explained the pattern of data observed, although it did not explain why a greater frequency of one type of offspring would be observed. Also notable was the fact that at the beginning of the unit 25% of the students indicated that it was possible for the offspring to receive genetic contributions from only one parent and not both (single characteristic category in Table 6). That is, they did not incorporate the notion that both parents contribute genetically to the offspring. By the end of the unit, however, no students were positing such an explanation. How did student ideas change? On the basis of a social constructivist framework, there were two mechanisms incorporated into the study to promote student construction of explanations. The first was that of providing anomalous instances of data as a catalyst for promoting cognitive disequilibrium. Students encountered these anomalous instances of data through their investigations with the fruit fly simulation software. To help students resolve that cognitive disequilibrium, an emphasis was placed on social construction of knowledge in which students working in pairs could coconstruct ideas, and students working with the teacher during whole-class discussion could develop explanations for anomalous patterns as a class. The emphasis was on a reflexive process of knowledge construction, in which individuals could contribute to whole-class discussion and whole-class discussion could contribute to individual ideas. This process also occurred with respect to guidance for approaches to scientific reasoning where exchanges with the dyads were then used to inform the whole class and vice versa. The role of instruction during this process was intended to guide and support discovery of anomalous patterns and development of explanations of those patterns. When students began to construct explanations for the anomalous data patterns, they did so in various ways. Mechanisms of chance, dominance, varying amounts, and varying strengths were all used as a means of explaining how the genetic contributions of the parents manifested themselves in the children. Students appeared to attend to various cues in the data, which may have triggered different schema to help them develop these explanations. For example, some students conceived of a fruit fly as always carrying both forms of the trait and that one appearance would occur over another by chance. This explanation resembled a coin-toss theory. That is, each fruit fly carried two forms of the trait; whichever form showed up was a matter of chance. Although this explanation is logical and neatly explains both patterns, it departs from the canonical notion, in that fruit flies do not always carry both forms of the trait, and both types of offspring do not always occur in equal amounts. Other students developed an intermediate model that incorporated the notion that a fruit fly could carry only one form of the trait or both forms of the trait; if both variations were present, then one variation would manifest

itself in the appearance of the offspring by chance. A more sophisticated variation of that model was that each fruit fly could carry one or both forms of the trait; if both variations were present, then the variation that was dominant dictated the appearance of the fruit fly. Students who posited varying amounts or strengths tended to refer to genetic material that could exist in varying amounts or strengths such as many genes, chromosomes, or varying amounts of blood or DNA. For the varying-amounts model, each variation of a trait could accumulate until there was “enough” of it to manifest itself in the offspring. For the varying-strengths model, the variations of a trait could be stronger or weaker, with the appearance of a particular variation indicating that the variation was strong enough to manifest itself. Although students had developed these models, they were not consistently used nor considered mutually exclusive. Specifically, approximately half of the students used either a dominance or chance explanation for one question and then a varying-amounts or strengths mechanism to explain the second question on the postassessment. In this sense, student responses resembled fledgling ideas that they had not yet consolidated into one consistent model. This type of response resembles that described as dual construction by Demastes et al. (1996), in which students apply logically incompatible conceptions. For example, describing genes as having varying strengths that would cause a particular appearance in the fruit fly is logically incompatible with then describing the appearance as being manifested by chance in another scenario. It may be that given more time and teacher support, students may have been able to generate more consistent models. Alternatively, providing direct instruction on particular aspects of the trait transmission process may have further facilitated student construction of ideas. Although approximately 80% of the students had developed some form of an explanation for at least one of the anomalous data patterns by the end of the unit, approximately 20% of the students had not (see Tables 4 and 5). Why were these students not able to construct an explanation? In order for conceptual change to occur, an assumption is made that an initial concept exists that can undergo change. These students may not have had an initial consistent model to explain the data. Therefore, further instances of anomalous data may have appeared equally inexplicable to them. Shepardson and Moje (1999) suggested that children with poorly developed understandings of phenomena are more likely to view anomalous data as irrelevant relative to those students with more well-developed understandings. Similar findings have been reported elsewhere regarding electric circuits (Schauble, Glaser, Raghavan, & Reiner, 1991) and genetics (Slack & Stewart, 1990), in which it has been suggested that a lack of domain-specific knowledge can hinder students in their ability to recognize and incorporate anomalous data. An alternative explanation is that an initial concept existed, but students may not have been able to construct an explanation that would allow them to make sense of the anomalous data. Therefore, these students may have simply ignored these data. For conceptual change to occur, anomalous data must be viewed as both intelligible and plausible (Posner et al., 1982). These data were likely intelligible but to render them plausible would have required a mechanism to explain them. Weak domain-specific knowledge may still have been the culprit in this case, in that students with weak genetics knowledge would be less capable of constructing a

KNOWLEDGE CONSTRUCTION AND SCIENTIFIC REASONING

373

genetic explanation relative to those with some basic genetics knowledge. Another finding that arose during the study was that more students developed an explanation for the mixed-breed anomalous outcome relative to the dominant anomalous outcome. The explanation that appeared to underlie this outcome was the fact that students observed the mixed-breed anomalous outcome significantly more often than the dominant outcome. Therefore, they had significantly more opportunities to theorize about the mixed-breed outcome relative to the dominant outcome. Anomalous outcomes that occur more frequently are less likely to be ignored or discounted and are also more likely to be construed as plausible rather than implausible (Koslowski, 1996). To consistently produce dominant hypotheses, students may have needed to encounter more instances of dominant outcomes. Students encountered twice as many mixed-breed anomalies as dominant anomalies. Not only were more mixed-breed anomalies encountered, they constituted a majority (62%) of the outcomes seen for the dominant– dominant test. Therefore, there may need to be a preponderance of evidence to the contrary in order for students to construct an explanation for an outcome that is at odds with an existing theory. In addition to students encountering fewer dominant anomalous outcomes, a complementary explanation is that perhaps it was more difficult, and potentially less intuitive, to propose a dominant pattern relative to a mixed-breed pattern. Specifically, it appeared that some students approached the investigation with the seemingly implicit assumption that both types of traits would behave in a similar manner. That is, students did not seem to assume that one trait should behave differently than another. This is a fair and logical assumption that would initially help to structure the data. In fact, this is the assumption of the null hypothesis encountered in social science research when comparing treatment groups: namely, that treatments are assumed to be equal until proven differently. Similarly, when students were confronted with two variations of the same trait, their initial hypothesis was that both variations would behave in a similar manner. This may have been the rationale underlying the notion that fruit flies always carried both forms of the trait and that whichever variation appeared was a matter of chance. Implicitly, each variation is equally likely to occur. In contrast, though, some students did recognize differences in the frequency with which each variation of the trait appeared. It is presumably these students who then constructed separate explanations for the dominant anomalous outcome. With respect to student scientific reasoning, students reacted to anomalies by proposing hypotheses and running tests in relation to the extent to which they encountered anomalous outcomes. Students ran significantly more tests for the more frequently occurring anomalous outcome relative to the test that produced no anomalous outcomes. They also generated more hypotheses for the more frequently occurring anomalous outcome. They ran fewer tests and proposed fewer anomalous hypotheses for the less frequently occurring anomalous outcome. When constructing their hypotheses, students seemed to vary in their tolerance for anomalous outcomes. One dyad seemed to switch from the standard hypotheses after one instance of anomalous data, whereas another dyad persevered with the standard hypotheses regardless of anomalous data encountered across several traits, until finally relenting and proposing one anomalous hypothesis.

Given that these students were novices in genetics, I expected they would evidence certain novice behaviors with respect to conducting unsystematic approaches to testing and conducting uncontrolled tests. In fact, some students did not initially test all possible combinations of tests, and approximately half of the dyads did not conduct a controlled test in which they could have varied one fruit fly at a time instead of mating two new fruit flies. However, by the end of the unit, 21 of 22 dyads comprehensively tested at least one of their traits, and 12 of 22 dyads conducted at least one controlled test. Further, some students were exhibiting quite sophisticated testing approaches by purposefully selecting fruit flies from vials that had a particular combination of parent characteristics, reflecting awareness of generational trends and incorporation of that information into their investigations. There also were changes in students’ causal, noncausal, and interaction inferences. At the beginning of the unit, some students had explained the anomalous inheritance patterns by indicating that the characteristics of only one parent had manifested itself in the offspring. The contributions of the other parent had gotten lost or waylaid in some manner. This is tantamount to an inference of inclusion for one parent and an inference of exclusion for the other parent. The actual canonical process involves an interaction between the genetic contributions of each parent. By the end of the unit, no students were positing an inference of inclusion for one parent and an inference of exclusion for the other parent. However, neither were all students positing interactions that incorporated the notion that both parents contributed to each fruit fly offspring. Instead, some students suggested that both parents contributed to the offspring but one parent contributed to some of the offspring and the other parent contributed to the rest of the offspring. This explanation was evident given the situation in which a red-eyed and white-eyed fruit fly were crossed to produce red-eyed and white-eyed offspring. In this case, some students suggested that the red-eyed parent contributed genetically to the red-eyed offspring and that the white-eyed parent contributed to the white-eyed offspring. This type of inference involves an inference of inclusion for each parent but not an interaction between both parent contributions for each child. It is not uncommon that this type of inference was somewhat elusive for these students, as interaction inferences are difficult for both children and adults to draw (Kuhn et al., 1995; Schauble, 1996). However, these shortcomings notwithstanding, it is important to note that these students were still able to gather data in a way that was sensible to them and allowed them to construct new ideas regarding trait transmission. In conclusion, the research presented here provided some insight about how students can react to anomalies in the context of an inquiry-oriented investigation. These findings are relevant for the practice of conducting inquiry-oriented investigations in the classroom as well as for expanding the knowledge base on responses to anomalous information by examining both knowledge construction and scientific reasoning together. Students generated hypotheses, ran tests, and constructed explanations in proportion to the extent to which they encountered anomalies. More anomalies meant that more hypotheses, tests, and explanations were generated. However, there may have been an interaction with the less frequently occurring anomaly also being less intuitive to explain. The more frequently occurring anomaly also occurred to an extremely high extent, comprising the majority of the outcomes for that test. Further research in other inquiry-oriented contexts is

374

ECHEVARRIA Limon, M., & Carretero, M. (1997). Conceptual change and anomalous data: A case study in the domain of natural sciences. European Journal of Psychology in Education, 12, 213–230. National Research Council. (1996). National science education standards. Washington, DC: National Academy Press. Piaget, J. (1963). The origins of intelligence in children (2nd ed.). London: Routledge & Kegan Paul. Posner, G. J., Strike, K. A., Hewson, P. W., & Gertzog, W. A. (1982). Accommodation of a scientific conception: Toward a theory of conceptual change. Science Education, 66, 211–227. Roth, W. M., & Roychoudhury, A. (1993). The development of science process skills in authentic contexts. Journal of Research in Science Teaching, 30, 127–152. Schauble, L. (1990). Belief revision in children: The role of prior knowledge and strategies for generating evidence. Journal of Experimental Child Psychology, 49, 31–57. Schauble, L. (1996). The development of scientific reasoning in knowledge-rich contexts. Developmental Psychology, 32, 102–119. Schauble, L., Glaser, R., Raghavan, K., & Reiner, M. (1991). Causal models and experimentation strategies in scientific reasoning. Journal of the Learning Sciences, 1, 201–238. Shepardson, D. P., & Moje, E. B. (1999). The role of anomalous data in restructuring fourth graders’ frameworks for understanding electric circuits. International Journal of Science Education, 21, 77–94. Simmons, P. E., & Lunetta, V. N. (1993). Problem-solving behaviors during a genetics computer simulation: Beyond the expert/novice dichotomy. Journal of Research in Science Teaching, 30, 153–173. Slack, S. J., & Stewart, J. (1990). High school students’ problem-solving performance on realistic genetics problems. Journal of Research in Science Teaching, 27, 55– 67. Smith, M. U., & Good, R. (1984). Problem solving and classical genetics: Successful versus unsuccessful performance. Journal of Research in Science Teaching, 21, 895–912. Stewart, J. (1983). Student problem solving in high school genetics. Science Education, 67, 523–540. Tao, P. K., & Gunstone, R. F. (1999). The process of conceptual change in force and motion during computer-supported physics instruction. Journal of Research in Science Teaching, 36, 859 – 882. Vollmeyer, R., Burns, B. D., & Holyoak, K. J. (1996). The impact of goal specificity on strategy use and the acquisition of problem structure. Cognitive Science, 20, 75–100. Vosniadou, S., & Brewer, W. (1987). Theories of knowledge restructuring in development. Review of Educational Research, 57, 51– 67. Vosniadou, S., & Brewer, W. F. (1992). Mental models of the Earth: A study of conceptual change in childhood. Cognitive Psychology, 24, 535–585.

needed to provide more information as to the consistency of the patterns presented here and to determine whether there are overall trends in approaches to scientific reasoning and knowledge construction that can be predicted when learners are presented with anomalous information.

References
Browning, M. E., & Lehman, J. D. (1988). Identification of student misconceptions in genetics problem solving via computer program. Journal of Research in Science Teaching, 25, 747–761. Burbules, N. C., & Linn, M. C. (1988). Response to contradiction: Scientific reasoning during adolescence. Journal of Educational Psychology, 80, 67–75. Carolina Biological Supply Company. (1989). Wisconsin Fast Plants: Investigating Mendelian genetics. Burlington, NC: Carolina Biological. Chinn, C., & Brewer, W. F. (1993). The role of anomalous data in knowledge acquisition: A theoretical framework and implications for science instruction. Review of Educational Research, 63, 1– 49. Cobb, P., & Yackel, E. (1995, October). Constructivist, emergent, and sociocultural perspectives in the context of developmental research. Paper presented at the North American Chapter of the International Group for the Psychology of Mathematics Education, Columbus, OH. Demastes, S. S., Good, R. G., & Peebles, P. (1996). Patterns of conceptual change in evolution. Journal of Research in Science Teaching, 33, 407– 431. Finkel, E. (1996). Making sense of genetics: Students’ knowledge use during problem solving in a high school genetics class. Journal of Research in Science Teaching, 33, 345–368. Friedler, Y., Nachmias, R., & Linn, M. C. (1990). Learning scientific reasoning skills in microcomputer-based laboratories. Journal of Research in Science Teaching, 27, 173–191. Graesser, A. C., & McMahen, C. L. (1993). Anomalous information triggers questions when adults solve quantitative problems and comprehend stories. Journal of Educational Psychology, 85, 136 –151. Hafner, R., & Stewart, J. (1995). Revising explanatory models to accommodate anomalous genetic phenomena: Problem solving in the “context of discovery.” Science Education, 79, 111–146. Jungck, J. R., & Calley, J. (1993). Genetics construction kit. In J. R. Jungck, P. Soderberg, J. Calley, N. Peterson, & J. Stewart (Eds.), The BioQUEST library. College Park: University of Maryland Press. Klahr, D., & Dunbar, K. (1988). Dual space search during scientific reasoning. Cognitive Science, 12, 1– 48. Klahr, D., Fay, A. L., & Dunbar, K. (1993). Heuristics for scientific experimentation: A developmental study. Cognitive Psychology, 25, 111–146. Koslowski, B. (1996). Theory and evidence: The development of scientific reasoning. Cambridge, MA: MIT Press. Kuhn, D., Garcia-Mila, M., Zohar, A., & Andersen, C. (1995). Strategies of knowledge acquisition. Monographs of the Society for Research in Child Development, 60(4, Serial No. 245).

Received November 30, 2001 Revision received March 12, 2002 Accepted March 14, 2002